17 Generative Adversarial Networks
Generative models use machine learning algorithms to learn a distribution from a set of training data and then generate new examples from that distribution. For example, a generative model might be trained on images of animals and then used to generate new images of animals. We can think of such a generative model in terms of a distribution p(x|w) in which x is a vector in the data space, and w represent the learnable parameters of the model. In many cases we are interested in conditional generative models of the form p(x|c, w) where c represents a vector of conditioning variables. In the case of our generative model for animal images, we may wish to specify that a generated image should be of a particular animal, such as a cat or a dog, specified by the value of c. For real-world applications such as image generation, the distributions are extremely complex, and consequently the introduction of deep learning has dramatically improved the performance of generative models. We have already encountered
© The Author(s), under exclusive license to Springer Nature Switzerland AG 2024 C. M. Bishop, H. Bishop, Deep Learning, https://doi.org/10.1007/978-3-031-45468-4_17
533